DeepMind is seeking an Expert Research Expert-Bench to design and implement multistep agentic tasks that simulate real-world research challenges. You will work in a tight loop with researchers, creating high-quality, hard tasks that require 1–2 days of effort across ML, data analysis, and software engineering.
The role emphasizes rigorous evaluation, Python scripting, and experience with notebook environments.
#J-18808-Ljbffr